The skinny on what's happening in AI - straight from the previous live session:
100 Issues! Wow! AND at the same time …
10,000 subscribers on Youtube:
and 10,000 followers on Instagram:
Thank you for your continued support. On all those platforms (and the 150,000 TikTok!).
It’s also a good moment to reflect on the most annoying aspect of all this. Anything worthwhile requires sustained effort. This has been 100 Lives, every morning getting up, reading the news and going Live. Then processing. Then publishing. Consistently, unerringly. Including whilst on “holiday”.
Consistency, not genius, is what matters. Just stubbornly turning up every day.
It’s boring!
Funnily enough over the weekend I saw this:
So, my question. And challenge. What’s the unreasonably boring thing you can just keep doing to outstrip everyone else?
As we move into Winter (well, at least us unlucky ones in the Northern hemisphere!) it’s super tempting to just stop. Rest. Come back in the New Year.
But this is exactly the time to go hard. Sure, to “get ahead” of the competition and all that. But more importantly for yourself. To prove to yourself that you are in it for the long haul.
Highlights:
📊 MIT Study Says AI Replaces 11.7% of Jobs (Spoiler: It Doesn't)
CNBC headline completely misrepresents MIT research. Study talks about task exposure, not job replacement. Journalists either can't read or won't. And then it gets further blown up on Twitter.
Kyle's take: This headline is bollocks. MIT's study found 11.7% of TASKS within jobs are technically exposed to AI, not that 11.7% of JOBS will be replaced. Massive difference.
Every job has multiple tasks - I talk to camera, set up lights, research articles, do bookkeeping. AI might handle some tasks but not the whole job.
The actual paper (which nobody reads because reading hard) specifically says "this is NOT about displacement" - it's about technical exposure of individual skills. CNBC either doesn't understand statistics or they're lying for clicks. I had to dig through MIT's website, create an account, download the PDF because journalists don't link sources anymore. They want you angry and sharing, not informed.
Source: MIT Iceberg Index / PDF study / CNBC coverage
💸 Deloitte Caught AGAIN: $1.6M Canadian Report Full of AI Hallucinations
Same company, different country. After Australia's AU$440k AI disaster, now Canada finds fake citations in healthcare report.
Kyle's take: Deloitte's at it again. First Australia (AU$440k report with made-up academics), now Canada ($1.6M healthcare report with fictional citations). Their response?
Basically: we stand by our recommendations, AI was only used in the bits you caught.
Absolute nonsense. They're charging millions in taxpayer money, running it through AI, not checking it, then gaslighting when caught.
This isn't just Deloitte being idiots - they're destroying the entire consulting industry. Why pay McKinsey millions when they're just using ChatGPT? The moat is gone. Companies will start going direct to AI.
The "cover your arse" insurance of hiring expensive consultants isn't worth the market’s billions anymore. IBM's already pivoting to implementation because they see the writing on the wall. Consulting as we know it is dead.
🦷 The Jagged Frontier: Why AI is Brilliant and Stupid Simultaneously
AI progress isn't uniform - it's jagged. Incredible at some tasks, useless at others.
Kyle's take: People think AI advances uniformly. Wrong. It's a jagged frontier - racing ahead in coding, stuck counting Rs in "strawberry."
This is why some people say "AI is amazing" while others say "it's useless" - depends which part of the frontier you're touching. We've poured resources into making AI code (engineers building for engineers) so it's brilliant there. But ask it to read the room in a meeting? Useless.
The frontier keeps expanding but unevenly. Artists went from "AI art is rubbish" to "AI is stealing our jobs" in two years as that frontier expanded in their domain. Ethan Mollick (who helped coined the term) says bottlenecks matter - if AI can't do ONE critical task in your workflow, the whole thing breaks. That's why lawyers still have jobs despite AI drafting contracts.
👀 Chinese Parents Deploy AI as Homework Prison Guards
ByteDance's Doubao AI watches kids study, yells when they slouch or check phones. Dystopian or brilliant? A little column A, a little column B…
Kyle's take: Chinese parents point camera at kid, tell AI "watch him study," and it becomes digital tiger parent. "Sit up straight! Don't chew your pen! Eyes away from phone!"
Funny video but massive opportunity here. Personal PT watching your form during home workouts? Posture monitor for desk-bound workers (I sat up straight just reading "your posture is off"!). Focus assistant for ADHD. So many applications.
This is using AI's vision capabilities for behavioural modification. I could build this in an afternoon - little camera, watches you work, keeps you accountable. People would pay for that. The building isn't the problem anymore - it's distribution and marketing! Too many opportunities, not enough time. Every funny AI video is a startup waiting to happen.
Member Questions:
"Absolutely rubbish at computers. What tools should I start with?" Asked at [00:24:45]
Kyle's response: Start with Lovable - looks like ChatGPT, non-intimidating, handles all backend stuff. Also Bolt, Base44, Replit, V0 but Lovable's most user-friendly. After few months, move to Cursor (what I use daily) or Claude Code. Personally I use Claude Code inside Cursor. Build tools for yourself first - I have dozens of vibe-coded tools on my site for SEO, newsletter processing. Start with personal tools before trying to sell products.
"How can we use AI agents to create blog posts that rank on Google?" Asked at [00:26:25]
Kyle's response: Careful - Google will deprioritise pure AI content. Start with HUMAN content. I talk for an hour, that becomes transcript, then newsletter, then expand into articles with AI. Pure ChatGPT articles are detectable and worthless. My process: Live show → YouTube → Transcript → Newsletter → Extract stories → Expand with AI. Creates 10+ pages daily that actually rank because they started human. Don't just tell ChatGPT "write article about X" - that's garbage nobody should read!
"ChatGPT 5.1 vs Claude vs Gemini as daily driver?" Asked at [00:31:00]
Kyle's response: Try them all. I use Claude for work (coding, brainstorming, courses) - I'm on $200/month plan Anthropic gave me so no limits. ChatGPT for personal stuff (travel, video games, recipes). Gemini for research - sometimes. If you're hitting Claude's limits (raised since 4.5 launched Monday), it becomes less useful. My split: Claude for serious work, ChatGPT for life stuff.
"From all your AI experience, how can someone start a business around AI?" Asked at [00:33:25]
Kyle's response: "Just doing AI" isn't a business. It's AI plus existing skills in market willing to pay. I built quiz for this exact question - https://aiwithkyle.com/tools/ai-business-quiz. Anyone saying "start AI trading bot" or "faceless YouTube" is selling you something. Depends on YOUR skills, time, whether it's side hustle or replacement. Check out my AI Business Quiz and 400,000+ words of free guides on the site.
Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.